Logic List Mailing Archive

Job Offers at the Intelligent Agents Lab Munich

The Intelligent Agents Lab - Call for Job Applications

http://www.ai-center.com/jobs/jobs.txt

(http://www.ai-center.com/jobs for the latest information and a Flash
version including further links.)


GOALS

- Real-Time Planning of an Intelligent Agent's Behavior -

The lab's goal is to develop and analyze theory and application of
intelligent agents, i.e., for the generation and execution of behavior
plans for goal-directed, autonomously acting units in realistic
environments including real time, dynamics, and open, not completely known
worlds.

Computer games will serve as primary application domain.

Experts from the areas of representation, reasoning and learning will
closely work together to realize - also practically - an agent
architecture as stated above. All research and code will be freely
available to the public. Starting point will be the EXCALIBUR agent
architecture.

- Technology Fusion -

For high-level topics such as intelligent agents, we need to bring many
technologies together. Some of the research areas involved have unluckily
already reached a counterproductive level of detail. A very high level of
specialization may make a difference for a specific application, but makes
it nearly impossible to integrate methods from other areas or to apply the
approaches to slightly different problems.

Our group will take one step back, and will work on optimizing the
interplay of all components instead of primarily focusing on single
techniques. In many cases, this will lead to radically different
approaches. Experts from the areas of representation, reasoning and
learning will closely work together on the same tasks at once, bringing in
the different perspectives, approaches and needs.

- Multi-Objective Algorithm Design -

Nearly all AI research focuses exclusively on increasing efficiency. Very
few research groups incorporate aspects such as real time, dynamics, an
open, not completely known world, and conceptional flexibility. It is
often argued that these "secondary" features could be considered later as
an add-on. However, similar to what has been stated in the previous
section, too much focus on one feature will hurt other features.
Trade-offs between speed and memory are well known, but the same applies
to other features: Algorithm design is a multi-objective task!

Because of the trade-offs involved in multi-objective tasks, neglecting
other important features leads to the somewhat counter-intuitive effect
that - although focusing on something obviously important like efficiency
- research steers away from possible applications, further widening the
gap between research and application. From the very first day on, our
group will thus pay attention to the complete set of important features,
optimizing the system from the holistic point of view, and not for single
properties in isolation.

- Computer Games (forget about chess) -

Computer games provide a perfect environment for our research because they
are a simulation of a highly complex and dynamic world, involving real
time, realistic physics, and so on. In contrast to the real world,
experiments can easily be controlled and repeated.

Besides serving as a complex simulator, computer games and virtual
environments are quickly gaining importance in general. According to a UK
government report, the computer games industry has even already become a
major contributor to the economy.

We are currently in contact with several game companies to evaluate
possible cooperations.

- Are you ready to push AI to the next level? -

If you have expertise in one or more of the following areas, we would be
happy to receive your application to join our team:

Intelligent agents, planning & scheduling, plan recognition, machine
learning & statistics, probability theory, knowledge representation,
decision theory, search control, constraint programming, local search

TEAM

We have job openings at all levels:
Post-Docs, PhD students and regular students

All positions start in January 2004.